EnterpriseDB
Staff Software Engineer, 2024 - Present
Depth
Working as a staff software engineer on a newly formed team to build a new AI PostgreSQL extension in Rust, I learned an incredible amount about database internals and machine learning applications. I authored the data preprocessing SDK that lets applications process unstructured data from native Postgres tables or any S3-compatible volume, eliminating separate ETL processes for operations like chunking, HTML parsing, and OCR. I also automated our GitHub CI/CD and caching for Rust testing and packaging, which made a real difference for code reliability and developer experience across the team.
Architecture
I built a model registry that allows users to call LLMs via SQL and developed an authentication framework using Postgres FDW user mappings to enable multi-user access to AI models. As these features grew, I architected a generic multistep Pipeline framework to orchestrate complex workflows and later, a dead letter queue system to log and retry errors in pipeline steps for better traceability. I also refactored Rust cloning, static dispatch, and take-by-value patterns to reduce the extension's memory usage.
Leadership
I mentored engineers on Rust performance and best practices and led a team to implement the generic pipeline step interface and multi-step orchestration APIs. It was amazing to present at community events to educate about using Rust to build performant applications and Postgres extensions. Meeting like-minded contributors and sharing my personal and professional projects was immensely fulfilling.